How Expert Tutors Revise Tutoring Policies and Strategies When Students Make Mistakes

This paper shows that expert tutors adjust both their high-level policies and their lower-level strategies, based on assessments of student performance. This contrasts with previous reports that tutors do not use such sophisticated methods in their tutoring. The specific strategy adjustments that we find are consistent with the interpretation that expert tutors seek to aid weaker students in practicing problem solving skills, perhaps to build a new mental model of the domain, and seek to aid stronger students in diagnosing and correcting errors in their current mental model. These results have implications for better understanding of human tutoring, as well as for the design of effective intelligent tutoring systems. We believe that computer tutors should be designed to use student assessments to determine how they tutor as well as what they tutor